Why We Forecast Ten Years, Not Thirty

Every CrestCast forecast runs ten years and then stops. There is no setting that extends it, and stated on its own like that the number reads like something we have not got round to yet.

It is the opposite. Ten years is a decision, and a fairly deliberate one, so it is worth explaining rather than leaving to look like a gap.

We could run thirty years, and we have removed every place that did

The forecast engine takes the number of years as an input. Ten is a value we hand it, not a wall it hits, and it will accept a good deal more than that. So the ten is not the engine's answer. It is ours.

The app used to let you choose. There was a picker in Settings offering five, ten, fifteen, twenty, twenty-five and thirty years, and we took it out in August 2026. There was also a life snapshot that ran thirty years through the same engine, and we have since removed that too. Deleting working features twice over is an odd thing to do, so here is the reasoning behind it.

A forecast stops being useful long before it stops being computable

A projection is only ever as good as what goes into it. Over a short distance that barely matters, because most of what happens in the next few years is already settled: the rate you are on, the fixed deal that ends in 2029, the nursery fees that turn into school fees, the loan with eleven payments left on it. Those are facts, and they have dates attached.

Stretch the same model across three decades and what it is made of changes completely. Very little of the answer is a fact any more. It is an inflation assumption, a growth assumption, a salary path and a household that does not exist yet, each one compounding on top of the last. The figure still arrives with a decimal point on it, and that precision has been borrowed from the arithmetic rather than earned by the inputs.

Here is what getting one assumption slightly wrong does. Take £3,000 a month of household spending and inflate it two ways: at 2.5%, which is the assumption CrestCast starts from, and at 3.5%. That is one percentage point, comfortably inside the range reasonable people disagree about.

At year ten the two lines sit about £390 a month apart, roughly 10%. That is close enough that the shape of the answer survives and you would make the same decision either way. At year thirty-five it is £7,120 against £10,001, a gap of about 40% on the same household doing exactly the same things.

The point is not that one of those lines is right. It is that at ten years the choice between them hardly changes what you should do, and at thirty-five years it changes everything, and nobody can tell you which to pick. Add a growth rate and a salary path on top, each with its own honest margin of error, and the spread gets wider again.

The assumptions are only half of the problem

The other half is that the plan itself expires. Somebody at thirty projecting to sixty-five is not modelling one life, they are modelling the first draft of several. Somewhere in those thirty-five years is a career change nobody has thought of yet, a house move, possibly a child, possibly a redundancy, an inheritance, a business, an illness. None of it is in the model, because none of it is knowable, and all of it matters more than the growth rate does.

That is not an argument for planning less. It is an argument that a thirty-five year projection is answering a question about a person who does not exist yet, and that the mistake is treating the answer as a plan. You would rebuild it from scratch several times before you ever arrived.

Ten years is about as far as a decision reaches

A more useful way to choose a horizon is to ask how far ahead the things you are actually deciding play out. For most households that lands at roughly a decade:

• A five year fixed rate, and the deal you remortgage onto after it
• A baby arriving, going through nursery, and reaching secondary school
• A car loan, a credit card and most personal debt clearing completely
• Whether to overpay the mortgage or invest, and long enough to see which one won
• Going part time, changing career, or one of you stopping work for a while

Every one of those is near enough to model in real detail, with real dates and your own numbers, and far enough out that you cannot hold it in your head. Stretching the same model to thirty-five years would not add a sixth decision to that list. It would put a much wider error bar around the five already on it.

When you genuinely do want thirty years

Sometimes you do, and the tools that offer it are not doing anything wrong. If the question is whether your pension gets you to retirement, a long projection is exactly the right shape of answer. It works by holding the assumptions still on purpose and reading the trend, which is the sensible move when the target is decades away and what you want to know is roughly whether you are pointed at it.

That is a different question from the one CrestCast answers rather than a longer version of it. A retirement projection is deliberately coarse across a long distance. A cashflow forecast is deliberately detailed across a short one. Asking either to do the other one's job is where the disappointment comes from.

Some tools do go further. PocketSmith, for one, tiers its horizon from six months up to sixty years on its top plan, and our comparison with PocketSmith sets out the rest of the differences.

So if what you want is a thirty year retirement projection, CrestCast is not that tool. It does not do retirement planning, and the forecast stops at ten years on purpose. What it does instead is answer the questions a retirement projection cannot: whether this decade's decisions work, month by month, in figures you can act on.

We had one long view left, and we have taken that out too

Until recently there was an exception. A life snapshot reached thirty years, showing roughly where the household landed and how old everyone was when it got there. It could carry the longer view because it answered a far coarser question than the forecast does: it never claimed to know your current account balance on a Tuesday in 2056.

It has now been removed. Keeping one thirty year figure in the corner of a product that argues against thirty year figures was a harder position to defend than simply not having it, and it was the last piece of the app pulling in the opposite direction to everything above. The principle it rested on is the one this whole piece rests on: the further ahead you look, the less detail you are entitled to claim, and the honest response to that is to stop rather than to blur.

If the broader question is which tool fits which job, where CrestCast sits between budgeting apps and retirement planning covers the same ground from the other direction.

And if you want the uncertainty argument in more depth, how accurate is a financial forecast goes through where the real error in a household projection actually comes from, which is rarely the arithmetic.